drivers/net/ibmveth.c

   1 /**************************************************************************/
   2 /*                                                                        */
   3 /* IBM eServer i/pSeries Virtual Ethernet Device Driver                   */
   4 /* Copyright (C) 2003 IBM Corp.                                           */
   5 /*  Originally written by Dave Larson (larson1@us.ibm.com)                */
   6 /*  Maintained by Santiago Leon (santil@us.ibm.com)                       */
   7 /*                                                                        */
   8 /*  This program is free software; you can redistribute it and/or modify  */
   9 /*  it under the terms of the GNU General Public License as published by  */
  10 /*  the Free Software Foundation; either version 2 of the License, or     */
  11 /*  (at your option) any later version.                                   */
  12 /*                                                                        */
  13 /*  This program is distributed in the hope that it will be useful,       */
  14 /*  but WITHOUT ANY WARRANTY; without even the implied warranty of        */
  15 /*  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         */
  16 /*  GNU General Public License for more details.                          */
  17 /*                                                                        */
  18 /*  You should have received a copy of the GNU General Public License     */
  19 /*  along with this program; if not, write to the Free Software           */
  20 /*  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  */
  21 /*                                                                   USA  */
  22 /*                                                                        */
  23 /* This module contains the implementation of a virtual ethernet device   */
  24 /* for use with IBM i/pSeries LPAR Linux.  It utilizes the logical LAN    */
  25 /* option of the RS/6000 Platform Architechture to interface with virtual */
  26 /* ethernet NICs that are presented to the partition by the hypervisor.   */
  27 /*                                                                        */
  28 /**************************************************************************/
  29 /*
  30   TODO:
  31   - add support for sysfs
  32   - possibly remove procfs support
  33 */
  34
  35 #include <linux/module.h>
  36 #include <linux/types.h>
  37 #include <linux/errno.h>
  38 #include <linux/ioport.h>
  39 #include <linux/dma-mapping.h>
  40 #include <linux/kernel.h>
  41 #include <linux/netdevice.h>
  42 #include <linux/etherdevice.h>
  43 #include <linux/skbuff.h>
  44 #include <linux/init.h>
  45 #include <linux/delay.h>
  46 #include <linux/mm.h>
  47 #include <linux/ethtool.h>
  48 #include <linux/proc_fs.h>
  49 #include <linux/in.h>
  50 #include <linux/ip.h>
  51 #include <net/net_namespace.h>
  52 #include <asm/hvcall.h>
  53 #include <asm/atomic.h>
  54 #include <asm/vio.h>
  55 #include <asm/uaccess.h>
  56 #include <linux/seq_file.h>
  57
  58 #include "ibmveth.h"
  59
  60 #undef DEBUG
  61
  62 #define ibmveth_printk(fmt, args...) \
  63   printk(KERN_DEBUG "%s: " fmt, __FILE__, ## args)
  64
  65 #define ibmveth_error_printk(fmt, args...) \
  66   printk(KERN_ERR "(%s:%3.3d ua:%x) ERROR: " fmt, __FILE__, __LINE__ , adapter->vdev->unit_address, ## args)
  67
  68 #ifdef DEBUG
  69 #define ibmveth_debug_printk_no_adapter(fmt, args...) \
  70   printk(KERN_DEBUG "(%s:%3.3d): " fmt, __FILE__, __LINE__ , ## args)
  71 #define ibmveth_debug_printk(fmt, args...) \
  72   printk(KERN_DEBUG "(%s:%3.3d ua:%x): " fmt, __FILE__, __LINE__ , adapter->vdev->unit_address, ## args)
  73 #define ibmveth_assert(expr) \
  74   if(!(expr)) {                                   \
  75     printk(KERN_DEBUG "assertion failed (%s:%3.3d ua:%x): %s\n", __FILE__, __LINE__, adapter->vdev->unit_address, #expr); \
  76     BUG(); \
  77   }
  78 #else
  79 #define ibmveth_debug_printk_no_adapter(fmt, args...)
  80 #define ibmveth_debug_printk(fmt, args...)
  81 #define ibmveth_assert(expr)
  82 #endif
  83
  84 static int ibmveth_open(struct net_device *dev);
  85 static int ibmveth_close(struct net_device *dev);
  86 static int ibmveth_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd);
  87 static int ibmveth_poll(struct napi_struct *napi, int budget);
  88 static int ibmveth_start_xmit(struct sk_buff *skb, struct net_device *dev);
  89 static void ibmveth_set_multicast_list(struct net_device *dev);
  90 static int ibmveth_change_mtu(struct net_device *dev, int new_mtu);
  91 static void ibmveth_proc_register_driver(void);
  92 static void ibmveth_proc_unregister_driver(void);
  93 static void ibmveth_proc_register_adapter(struct ibmveth_adapter *adapter);
  94 static void ibmveth_proc_unregister_adapter(struct ibmveth_adapter *adapter);
  95 static irqreturn_t ibmveth_interrupt(int irq, void *dev_instance);
  96 static void ibmveth_rxq_harvest_buffer(struct ibmveth_adapter *adapter);
  97 static struct kobj_type ktype_veth_pool;
  98
  99 #ifdef CONFIG_PROC_FS
 100 #define IBMVETH_PROC_DIR "ibmveth"
 101 static struct proc_dir_entry *ibmveth_proc_dir;
 102 #endif
 103
 104 static const char ibmveth_driver_name[] = "ibmveth";
 105 static const char ibmveth_driver_string[] = "IBM i/pSeries Virtual Ethernet Driver";
 106 #define ibmveth_driver_version "1.03"
 107
 108 MODULE_AUTHOR("Santiago Leon <santil@us.ibm.com>");
 109 MODULE_DESCRIPTION("IBM i/pSeries Virtual Ethernet Driver");
 110 MODULE_LICENSE("GPL");
 111 MODULE_VERSION(ibmveth_driver_version);
 112
 113 struct ibmveth_stat {
 114         char name[ETH_GSTRING_LEN];
 115         int offset;
 116 };
 117
 118 #define IBMVETH_STAT_OFF(stat) offsetof(struct ibmveth_adapter, stat)
 119 #define IBMVETH_GET_STAT(a, off) *((u64 *)(((unsigned long)(a)) + off))
 120
 121 struct ibmveth_stat ibmveth_stats[] = {
 122         { "replenish_task_cycles", IBMVETH_STAT_OFF(replenish_task_cycles) },
 123         { "replenish_no_mem", IBMVETH_STAT_OFF(replenish_no_mem) },
 124         { "replenish_add_buff_failure", IBMVETH_STAT_OFF(replenish_add_buff_failure) },
 125         { "replenish_add_buff_success", IBMVETH_STAT_OFF(replenish_add_buff_success) },
 126         { "rx_invalid_buffer", IBMVETH_STAT_OFF(rx_invalid_buffer) },
 127         { "rx_no_buffer", IBMVETH_STAT_OFF(rx_no_buffer) },
 128         { "tx_map_failed", IBMVETH_STAT_OFF(tx_map_failed) },
 129         { "tx_send_failed", IBMVETH_STAT_OFF(tx_send_failed) },
 130 };
 131
 132 /* simple methods of getting data from the current rxq entry */
 133 static inline u32 ibmveth_rxq_flags(struct ibmveth_adapter *adapter)
 134 {
 135         return adapter->rx_queue.queue_addr[adapter->rx_queue.index].flags_off;
 136 }
 137
 138 static inline int ibmveth_rxq_toggle(struct ibmveth_adapter *adapter)
 139 {
 140         return (ibmveth_rxq_flags(adapter) & IBMVETH_RXQ_TOGGLE) >> IBMVETH_RXQ_TOGGLE_SHIFT;
 141 }
 142
 143 static inline int ibmveth_rxq_pending_buffer(struct ibmveth_adapter *adapter)
 144 {
 145         return (ibmveth_rxq_toggle(adapter) == adapter->rx_queue.toggle);
 146 }
 147
 148 static inline int ibmveth_rxq_buffer_valid(struct ibmveth_adapter *adapter)
 149 {
 150         return (ibmveth_rxq_flags(adapter) & IBMVETH_RXQ_VALID);
 151 }
 152
 153 static inline int ibmveth_rxq_frame_offset(struct ibmveth_adapter *adapter)
 154 {
 155         return (ibmveth_rxq_flags(adapter) & IBMVETH_RXQ_OFF_MASK);
 156 }
 157
 158 static inline int ibmveth_rxq_frame_length(struct ibmveth_adapter *adapter)
 159 {
 160         return (adapter->rx_queue.queue_addr[adapter->rx_queue.index].length);
 161 }
 162
 163 static inline int ibmveth_rxq_csum_good(struct ibmveth_adapter *adapter)
 164 {
 165         return (ibmveth_rxq_flags(adapter) & IBMVETH_RXQ_CSUM_GOOD);
 166 }
 167
 168 /* setup the initial settings for a buffer pool */
 169 static void ibmveth_init_buffer_pool(struct ibmveth_buff_pool *pool, u32 pool_index, u32 pool_size, u32 buff_size, u32 pool_active)
 170 {
 171         pool->size = pool_size;
 172         pool->index = pool_index;
 173         pool->buff_size = buff_size;
 174         pool->threshold = pool_size / 2;
 175         pool->active = pool_active;
 176 }
 177
 178 /* allocate and setup an buffer pool - called during open */
 179 static int ibmveth_alloc_buffer_pool(struct ibmveth_buff_pool *pool)
 180 {
 181         int i;
 182
 183         pool->free_map = kmalloc(sizeof(u16) * pool->size, GFP_KERNEL);
 184
 185         if(!pool->free_map) {
 186                 return -1;
 187         }
 188
 189         pool->dma_addr = kmalloc(sizeof(dma_addr_t) * pool->size, GFP_KERNEL);
 190         if(!pool->dma_addr) {
 191                 kfree(pool->free_map);
 192                 pool->free_map = NULL;
 193                 return -1;
 194         }
 195
 196         pool->skbuff = kmalloc(sizeof(void*) * pool->size, GFP_KERNEL);
 197
 198         if(!pool->skbuff) {
 199                 kfree(pool->dma_addr);
 200                 pool->dma_addr = NULL;
 201
 202                 kfree(pool->free_map);
 203                 pool->free_map = NULL;
 204                 return -1;
 205         }
 206
 207         memset(pool->skbuff, 0, sizeof(void*) * pool->size);
 208         memset(pool->dma_addr, 0, sizeof(dma_addr_t) * pool->size);
 209
 210         for(i = 0; i < pool->size; ++i) {
 211                 pool->free_map[i] = i;
 212         }
 213
 214         atomic_set(&pool->available, 0);
 215         pool->producer_index = 0;
 216         pool->consumer_index = 0;
 217
 218         return 0;
 219 }
 220
 221 /* replenish the buffers for a pool.  note that we don't need to
 222  * skb_reserve these since they are used for incoming...
 223  */
 224 static void ibmveth_replenish_buffer_pool(struct ibmveth_adapter *adapter, struct ibmveth_buff_pool *pool)
 225 {
 226         u32 i;
 227         u32 count = pool->size - atomic_read(&pool->available);
 228         u32 buffers_added = 0;
 229
 230         mb();
 231
 232         for(i = 0; i < count; ++i) {
 233                 struct sk_buff *skb;
 234                 unsigned int free_index, index;
 235                 u64 correlator;
 236                 union ibmveth_buf_desc desc;
 237                 unsigned long lpar_rc;
 238                 dma_addr_t dma_addr;
 239
 240                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 241
 242                 if(!skb) {
 243                         ibmveth_debug_printk("replenish: unable to allocate skb\n");
 244                         adapter->replenish_no_mem++;
 245                         break;
 246                 }
 247
 248                 free_index = pool->consumer_index;
 249                 pool->consumer_index = (pool->consumer_index + 1) % pool->size;
 250                 index = pool->free_map[free_index];
 251
 252                 ibmveth_assert(index != IBM_VETH_INVALID_MAP);
 253                 ibmveth_assert(pool->skbuff[index] == NULL);
 254
 255                 dma_addr = dma_map_single(&adapter->vdev->dev, skb->data,
 256                                 pool->buff_size, DMA_FROM_DEVICE);
 257
 258                 pool->free_map[free_index] = IBM_VETH_INVALID_MAP;
 259                 pool->dma_addr[index] = dma_addr;
 260                 pool->skbuff[index] = skb;
 261
 262                 correlator = ((u64)pool->index << 32) | index;
 263                 *(u64*)skb->data = correlator;
 264
 265                 desc.fields.flags_len = IBMVETH_BUF_VALID | pool->buff_size;
 266                 desc.fields.address = dma_addr;
 267
 268                 lpar_rc = h_add_logical_lan_buffer(adapter->vdev->unit_address, desc.desc);
 269
 270                 if(lpar_rc != H_SUCCESS) {
 271                         pool->free_map[free_index] = index;
 272                         pool->skbuff[index] = NULL;
 273                         if (pool->consumer_index == 0)
 274                                 pool->consumer_index = pool->size - 1;
 275                         else
 276                                 pool->consumer_index--;
 277                         dma_unmap_single(&adapter->vdev->dev,
 278                                         pool->dma_addr[index], pool->buff_size,
 279                                         DMA_FROM_DEVICE);
 280                         dev_kfree_skb_any(skb);
 281                         adapter->replenish_add_buff_failure++;
 282                         break;
 283                 } else {
 284                         buffers_added++;
 285                         adapter->replenish_add_buff_success++;
 286                 }
 287         }
 288
 289         mb();
 290         atomic_add(buffers_added, &(pool->available));
 291 }
 292
 293 /* replenish routine */
 294 static void ibmveth_replenish_task(struct ibmveth_adapter *adapter)
 295 {
 296         int i;
 297
 298         adapter->replenish_task_cycles++;
 299
 300         for(i = 0; i < IbmVethNumBufferPools; i++)
 301                 if(adapter->rx_buff_pool[i].active)
 302                         ibmveth_replenish_buffer_pool(adapter,
 303                                                      &adapter->rx_buff_pool[i]);
 304
 305         adapter->rx_no_buffer = *(u64*)(((char*)adapter->buffer_list_addr) + 4096 - 8);
 306 }
 307
 308 /* empty and free ana buffer pool - also used to do cleanup in error paths */
 309 static void ibmveth_free_buffer_pool(struct ibmveth_adapter *adapter, struct ibmveth_buff_pool *pool)
 310 {
 311         int i;
 312
 313         kfree(pool->free_map);
 314         pool->free_map = NULL;
 315
 316         if(pool->skbuff && pool->dma_addr) {
 317                 for(i = 0; i < pool->size; ++i) {
 318                         struct sk_buff *skb = pool->skbuff[i];
 319                         if(skb) {
 320                                 dma_unmap_single(&adapter->vdev->dev,
 321                                                  pool->dma_addr[i],
 322                                                  pool->buff_size,
 323                                                  DMA_FROM_DEVICE);
 324                                 dev_kfree_skb_any(skb);
 325                                 pool->skbuff[i] = NULL;
 326                         }
 327                 }
 328         }
 329
 330         if(pool->dma_addr) {
 331                 kfree(pool->dma_addr);
 332                 pool->dma_addr = NULL;
 333         }
 334
 335         if(pool->skbuff) {
 336                 kfree(pool->skbuff);
 337                 pool->skbuff = NULL;
 338         }
 339 }
 340
 341 /* remove a buffer from a pool */
 342 static void ibmveth_remove_buffer_from_pool(struct ibmveth_adapter *adapter, u64 correlator)
 343 {
 344         unsigned int pool  = correlator >> 32;
 345         unsigned int index = correlator & 0xffffffffUL;
 346         unsigned int free_index;
 347         struct sk_buff *skb;
 348
 349         ibmveth_assert(pool < IbmVethNumBufferPools);
 350         ibmveth_assert(index < adapter->rx_buff_pool[pool].size);
 351
 352         skb = adapter->rx_buff_pool[pool].skbuff[index];
 353
 354         ibmveth_assert(skb != NULL);
 355
 356         adapter->rx_buff_pool[pool].skbuff[index] = NULL;
 357
 358         dma_unmap_single(&adapter->vdev->dev,
 359                          adapter->rx_buff_pool[pool].dma_addr[index],
 360                          adapter->rx_buff_pool[pool].buff_size,
 361                          DMA_FROM_DEVICE);
 362
 363         free_index = adapter->rx_buff_pool[pool].producer_index;
 364         adapter->rx_buff_pool[pool].producer_index
 365                 = (adapter->rx_buff_pool[pool].producer_index + 1)
 366                 % adapter->rx_buff_pool[pool].size;
 367         adapter->rx_buff_pool[pool].free_map[free_index] = index;
 368
 369         mb();
 370
 371         atomic_dec(&(adapter->rx_buff_pool[pool].available));
 372 }
 373
 374 /* get the current buffer on the rx queue */
 375 static inline struct sk_buff *ibmveth_rxq_get_buffer(struct ibmveth_adapter *adapter)
 376 {
 377         u64 correlator = adapter->rx_queue.queue_addr[adapter->rx_queue.index].correlator;
 378         unsigned int pool = correlator >> 32;
 379         unsigned int index = correlator & 0xffffffffUL;
 380
 381         ibmveth_assert(pool < IbmVethNumBufferPools);
 382         ibmveth_assert(index < adapter->rx_buff_pool[pool].size);
 383
 384         return adapter->rx_buff_pool[pool].skbuff[index];
 385 }
 386
 387 /* recycle the current buffer on the rx queue */
 388 static void ibmveth_rxq_recycle_buffer(struct ibmveth_adapter *adapter)
 389 {
 390         u32 q_index = adapter->rx_queue.index;
 391         u64 correlator = adapter->rx_queue.queue_addr[q_index].correlator;
 392         unsigned int pool = correlator >> 32;
 393         unsigned int index = correlator & 0xffffffffUL;
 394         union ibmveth_buf_desc desc;
 395         unsigned long lpar_rc;
 396
 397         ibmveth_assert(pool < IbmVethNumBufferPools);
 398         ibmveth_assert(index < adapter->rx_buff_pool[pool].size);
 399
 400         if(!adapter->rx_buff_pool[pool].active) {
 401                 ibmveth_rxq_harvest_buffer(adapter);
 402                 ibmveth_free_buffer_pool(adapter, &adapter->rx_buff_pool[pool]);
 403                 return;
 404         }
 405
 406         desc.fields.flags_len = IBMVETH_BUF_VALID |
 407                 adapter->rx_buff_pool[pool].buff_size;
 408         desc.fields.address = adapter->rx_buff_pool[pool].dma_addr[index];
 409
 410         lpar_rc = h_add_logical_lan_buffer(adapter->vdev->unit_address, desc.desc);
 411
 412         if(lpar_rc != H_SUCCESS) {
 413                 ibmveth_debug_printk("h_add_logical_lan_buffer failed during recycle rc=%ld", lpar_rc);
 414                 ibmveth_remove_buffer_from_pool(adapter, adapter->rx_queue.queue_addr[adapter->rx_queue.index].correlator);
 415         }
 416
 417         if(++adapter->rx_queue.index == adapter->rx_queue.num_slots) {
 418                 adapter->rx_queue.index = 0;
 419                 adapter->rx_queue.toggle = !adapter->rx_queue.toggle;
 420         }
 421 }
 422
 423 static void ibmveth_rxq_harvest_buffer(struct ibmveth_adapter *adapter)
 424 {
 425         ibmveth_remove_buffer_from_pool(adapter, adapter->rx_queue.queue_addr[adapter->rx_queue.index].correlator);
 426
 427         if(++adapter->rx_queue.index == adapter->rx_queue.num_slots) {
 428                 adapter->rx_queue.index = 0;
 429                 adapter->rx_queue.toggle = !adapter->rx_queue.toggle;
 430         }
 431 }
 432
 433 static void ibmveth_cleanup(struct ibmveth_adapter *adapter)
 434 {
 435         int i;
 436
 437         if(adapter->buffer_list_addr != NULL) {
 438                 if(!dma_mapping_error(adapter->buffer_list_dma)) {
 439                         dma_unmap_single(&adapter->vdev->dev,
 440                                         adapter->buffer_list_dma, 4096,
 441                                         DMA_BIDIRECTIONAL);
 442                         adapter->buffer_list_dma = DMA_ERROR_CODE;
 443                 }
 444                 free_page((unsigned long)adapter->buffer_list_addr);
 445                 adapter->buffer_list_addr = NULL;
 446         }
 447
 448         if(adapter->filter_list_addr != NULL) {
 449                 if(!dma_mapping_error(adapter->filter_list_dma)) {
 450                         dma_unmap_single(&adapter->vdev->dev,
 451                                         adapter->filter_list_dma, 4096,
 452                                         DMA_BIDIRECTIONAL);
 453                         adapter->filter_list_dma = DMA_ERROR_CODE;
 454                 }
 455                 free_page((unsigned long)adapter->filter_list_addr);
 456                 adapter->filter_list_addr = NULL;
 457         }
 458
 459         if(adapter->rx_queue.queue_addr != NULL) {
 460                 if(!dma_mapping_error(adapter->rx_queue.queue_dma)) {
 461                         dma_unmap_single(&adapter->vdev->dev,
 462                                         adapter->rx_queue.queue_dma,
 463                                         adapter->rx_queue.queue_len,
 464                                         DMA_BIDIRECTIONAL);
 465                         adapter->rx_queue.queue_dma = DMA_ERROR_CODE;
 466                 }
 467                 kfree(adapter->rx_queue.queue_addr);
 468                 adapter->rx_queue.queue_addr = NULL;
 469         }
 470
 471         for(i = 0; i<IbmVethNumBufferPools; i++)
 472                 if (adapter->rx_buff_pool[i].active)
 473                         ibmveth_free_buffer_pool(adapter,
 474                                                  &adapter->rx_buff_pool[i]);
 475 }
 476
 477 static int ibmveth_register_logical_lan(struct ibmveth_adapter *adapter,
 478         union ibmveth_buf_desc rxq_desc, u64 mac_address)
 479 {
 480         int rc, try_again = 1;
 481
 482         /* After a kexec the adapter will still be open, so our attempt to
 483         * open it will fail. So if we get a failure we free the adapter and
 484         * try again, but only once. */
 485 retry:
 486         rc = h_register_logical_lan(adapter->vdev->unit_address,
 487                                     adapter->buffer_list_dma, rxq_desc.desc,
 488                                     adapter->filter_list_dma, mac_address);
 489
 490         if (rc != H_SUCCESS && try_again) {
 491                 do {
 492                         rc = h_free_logical_lan(adapter->vdev->unit_address);
 493                 } while (H_IS_LONG_BUSY(rc) || (rc == H_BUSY));
 494
 495                 try_again = 0;
 496                 goto retry;
 497         }
 498
 499         return rc;
 500 }
 501
 502 static int ibmveth_open(struct net_device *netdev)
 503 {
 504         struct ibmveth_adapter *adapter = netdev->priv;
 505         u64 mac_address = 0;
 506         int rxq_entries = 1;
 507         unsigned long lpar_rc;
 508         int rc;
 509         union ibmveth_buf_desc rxq_desc;
 510         int i;
 511
 512         ibmveth_debug_printk("open starting\n");
 513
 514         napi_enable(&adapter->napi);
 515
 516         for(i = 0; i<IbmVethNumBufferPools; i++)
 517                 rxq_entries += adapter->rx_buff_pool[i].size;
 518
 519         adapter->buffer_list_addr = (void*) get_zeroed_page(GFP_KERNEL);
 520         adapter->filter_list_addr = (void*) get_zeroed_page(GFP_KERNEL);
 521
 522         if(!adapter->buffer_list_addr || !adapter->filter_list_addr) {
 523                 ibmveth_error_printk("unable to allocate filter or buffer list pages\n");
 524                 ibmveth_cleanup(adapter);
 525                 napi_disable(&adapter->napi);
 526                 return -ENOMEM;
 527         }
 528
 529         adapter->rx_queue.queue_len = sizeof(struct ibmveth_rx_q_entry) * rxq_entries;
 530         adapter->rx_queue.queue_addr = kmalloc(adapter->rx_queue.queue_len, GFP_KERNEL);
 531
 532         if(!adapter->rx_queue.queue_addr) {
 533                 ibmveth_error_printk("unable to allocate rx queue pages\n");
 534                 ibmveth_cleanup(adapter);
 535                 napi_disable(&adapter->napi);
 536                 return -ENOMEM;
 537         }
 538
 539         adapter->buffer_list_dma = dma_map_single(&adapter->vdev->dev,
 540                         adapter->buffer_list_addr, 4096, DMA_BIDIRECTIONAL);
 541         adapter->filter_list_dma = dma_map_single(&adapter->vdev->dev,
 542                         adapter->filter_list_addr, 4096, DMA_BIDIRECTIONAL);
 543         adapter->rx_queue.queue_dma = dma_map_single(&adapter->vdev->dev,
 544                         adapter->rx_queue.queue_addr,
 545                         adapter->rx_queue.queue_len, DMA_BIDIRECTIONAL);
 546
 547         if((dma_mapping_error(adapter->buffer_list_dma) ) ||
 548            (dma_mapping_error(adapter->filter_list_dma)) ||
 549            (dma_mapping_error(adapter->rx_queue.queue_dma))) {
 550                 ibmveth_error_printk("unable to map filter or buffer list pages\n");
 551                 ibmveth_cleanup(adapter);
 552                 napi_disable(&adapter->napi);
 553                 return -ENOMEM;
 554         }
 555
 556         adapter->rx_queue.index = 0;
 557         adapter->rx_queue.num_slots = rxq_entries;
 558         adapter->rx_queue.toggle = 1;
 559
 560         memcpy(&mac_address, netdev->dev_addr, netdev->addr_len);
 561         mac_address = mac_address >> 16;
 562
 563         rxq_desc.fields.flags_len = IBMVETH_BUF_VALID | adapter->rx_queue.queue_len;
 564         rxq_desc.fields.address = adapter->rx_queue.queue_dma;
 565
 566         ibmveth_debug_printk("buffer list @ 0x%p\n", adapter->buffer_list_addr);
 567         ibmveth_debug_printk("filter list @ 0x%p\n", adapter->filter_list_addr);
 568         ibmveth_debug_printk("receive q   @ 0x%p\n", adapter->rx_queue.queue_addr);
 569
 570         h_vio_signal(adapter->vdev->unit_address, VIO_IRQ_DISABLE);
 571
 572         lpar_rc = ibmveth_register_logical_lan(adapter, rxq_desc, mac_address);
 573
 574         if(lpar_rc != H_SUCCESS) {
 575                 ibmveth_error_printk("h_register_logical_lan failed with %ld\n", lpar_rc);
 576                 ibmveth_error_printk("buffer TCE:0x%lx filter TCE:0x%lx rxq desc:0x%lx MAC:0x%lx\n",
 577                                      adapter->buffer_list_dma,
 578                                      adapter->filter_list_dma,
 579                                      rxq_desc.desc,
 580                                      mac_address);
 581                 ibmveth_cleanup(adapter);
 582                 napi_disable(&adapter->napi);
 583                 return -ENONET;
 584         }
 585
 586         for(i = 0; i<IbmVethNumBufferPools; i++) {
 587                 if(!adapter->rx_buff_pool[i].active)
 588                         continue;
 589                 if (ibmveth_alloc_buffer_pool(&adapter->rx_buff_pool[i])) {
 590                         ibmveth_error_printk("unable to alloc pool\n");
 591                         adapter->rx_buff_pool[i].active = 0;
 592                         ibmveth_cleanup(adapter);
 593                         napi_disable(&adapter->napi);
 594                         return -ENOMEM ;
 595                 }
 596         }
 597
 598         ibmveth_debug_printk("registering irq 0x%x\n", netdev->irq);
 599         if((rc = request_irq(netdev->irq, &ibmveth_interrupt, 0, netdev->name, netdev)) != 0) {
 600                 ibmveth_error_printk("unable to request irq 0x%x, rc %d\n", netdev->irq, rc);
 601                 do {
 602                         rc = h_free_logical_lan(adapter->vdev->unit_address);
 603                 } while (H_IS_LONG_BUSY(rc) || (rc == H_BUSY));
 604
 605                 ibmveth_cleanup(adapter);
 606                 napi_disable(&adapter->napi);
 607                 return rc;
 608         }
 609
 610         ibmveth_debug_printk("initial replenish cycle\n");
 611         ibmveth_interrupt(netdev->irq, netdev);
 612
 613         netif_start_queue(netdev);
 614
 615         ibmveth_debug_printk("open complete\n");
 616
 617         return 0;
 618 }
 619
 620 static int ibmveth_close(struct net_device *netdev)
 621 {
 622         struct ibmveth_adapter *adapter = netdev->priv;
 623         long lpar_rc;
 624
 625         ibmveth_debug_printk("close starting\n");
 626
 627         napi_disable(&adapter->napi);
 628
 629         if (!adapter->pool_config)
 630                 netif_stop_queue(netdev);
 631
 632         free_irq(netdev->irq, netdev);
 633
 634         do {
 635                 lpar_rc = h_free_logical_lan(adapter->vdev->unit_address);
 636         } while (H_IS_LONG_BUSY(lpar_rc) || (lpar_rc == H_BUSY));
 637
 638         if(lpar_rc != H_SUCCESS)
 639         {
 640                 ibmveth_error_printk("h_free_logical_lan failed with %lx, continuing with close\n",
 641                                      lpar_rc);
 642         }
 643
 644         adapter->rx_no_buffer = *(u64*)(((char*)adapter->buffer_list_addr) + 4096 - 8);
 645
 646         ibmveth_cleanup(adapter);
 647
 648         ibmveth_debug_printk("close complete\n");
 649
 650         return 0;
 651 }
 652
 653 static int netdev_get_settings(struct net_device *dev, struct ethtool_cmd *cmd) {
 654         cmd->supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg | SUPPORTED_FIBRE);
 655         cmd->advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg | ADVERTISED_FIBRE);
 656         cmd->speed = SPEED_1000;
 657         cmd->duplex = DUPLEX_FULL;
 658         cmd->port = PORT_FIBRE;
 659         cmd->phy_address = 0;
 660         cmd->transceiver = XCVR_INTERNAL;
 661         cmd->autoneg = AUTONEG_ENABLE;
 662         cmd->maxtxpkt = 0;
 663         cmd->maxrxpkt = 1;
 664         return 0;
 665 }
 666
 667 static void netdev_get_drvinfo (struct net_device *dev, struct ethtool_drvinfo *info) {
 668         strncpy(info->driver, ibmveth_driver_name, sizeof(info->driver) - 1);
 669         strncpy(info->version, ibmveth_driver_version, sizeof(info->version) - 1);
 670 }
 671
 672 static u32 netdev_get_link(struct net_device *dev) {
 673         return 1;
 674 }
 675
 676 static void ibmveth_set_rx_csum_flags(struct net_device *dev, u32 data)
 677 {
 678         struct ibmveth_adapter *adapter = dev->priv;
 679
 680         if (data)
 681                 adapter->rx_csum = 1;
 682         else {
 683                 /*
 684                  * Since the ibmveth firmware interface does not have the concept of
 685                  * separate tx/rx checksum offload enable, if rx checksum is disabled
 686                  * we also have to disable tx checksum offload. Once we disable rx
 687                  * checksum offload, we are no longer allowed to send tx buffers that
 688                  * are not properly checksummed.
 689                  */
 690                 adapter->rx_csum = 0;
 691                 dev->features &= ~NETIF_F_IP_CSUM;
 692         }
 693 }
 694
 695 static void ibmveth_set_tx_csum_flags(struct net_device *dev, u32 data)
 696 {
 697         struct ibmveth_adapter *adapter = dev->priv;
 698
 699         if (data) {
 700                 dev->features |= NETIF_F_IP_CSUM;
 701                 adapter->rx_csum = 1;
 702         } else
 703                 dev->features &= ~NETIF_F_IP_CSUM;
 704 }
 705
 706 static int ibmveth_set_csum_offload(struct net_device *dev, u32 data,
 707                                     void (*done) (struct net_device *, u32))
 708 {
 709         struct ibmveth_adapter *adapter = dev->priv;
 710         u64 set_attr, clr_attr, ret_attr;
 711         long ret;
 712         int rc1 = 0, rc2 = 0;
 713         int restart = 0;
 714
 715         if (netif_running(dev)) {
 716                 restart = 1;
 717                 adapter->pool_config = 1;
 718                 ibmveth_close(dev);
 719                 adapter->pool_config = 0;
 720         }
 721
 722         set_attr = 0;
 723         clr_attr = 0;
 724
 725         if (data)
 726                 set_attr = IBMVETH_ILLAN_IPV4_TCP_CSUM;
 727         else
 728                 clr_attr = IBMVETH_ILLAN_IPV4_TCP_CSUM;
 729
 730         ret = h_illan_attributes(adapter->vdev->unit_address, 0, 0, &ret_attr);
 731
 732         if (ret == H_SUCCESS && !(ret_attr & IBMVETH_ILLAN_ACTIVE_TRUNK) &&
 733             !(ret_attr & IBMVETH_ILLAN_TRUNK_PRI_MASK) &&
 734             (ret_attr & IBMVETH_ILLAN_PADDED_PKT_CSUM)) {
 735                 ret = h_illan_attributes(adapter->vdev->unit_address, clr_attr,
 736                                          set_attr, &ret_attr);
 737
 738                 if (ret != H_SUCCESS) {
 739                         rc1 = -EIO;
 740                         ibmveth_error_printk("unable to change checksum offload settings."
 741                                              " %d rc=%ld\n", data, ret);
 742
 743                         ret = h_illan_attributes(adapter->vdev->unit_address,
 744                                                  set_attr, clr_attr, &ret_attr);
 745                 } else
 746                         done(dev, data);
 747         } else {
 748                 rc1 = -EIO;
 749                 ibmveth_error_printk("unable to change checksum offload settings."
 750                                      " %d rc=%ld ret_attr=%lx\n", data, ret, ret_attr);
 751         }
 752
 753         if (restart)
 754                 rc2 = ibmveth_open(dev);
 755
 756         return rc1 ? rc1 : rc2;
 757 }
 758
 759 static int ibmveth_set_rx_csum(struct net_device *dev, u32 data)
 760 {
 761         struct ibmveth_adapter *adapter = dev->priv;
 762
 763         if ((data && adapter->rx_csum) || (!data && !adapter->rx_csum))
 764                 return 0;
 765
 766         return ibmveth_set_csum_offload(dev, data, ibmveth_set_rx_csum_flags);
 767 }
 768
 769 static int ibmveth_set_tx_csum(struct net_device *dev, u32 data)
 770 {
 771         struct ibmveth_adapter *adapter = dev->priv;
 772         int rc = 0;
 773
 774         if (data && (dev->features & NETIF_F_IP_CSUM))
 775                 return 0;
 776         if (!data && !(dev->features & NETIF_F_IP_CSUM))
 777                 return 0;
 778
 779         if (data && !adapter->rx_csum)
 780                 rc = ibmveth_set_csum_offload(dev, data, ibmveth_set_tx_csum_flags);
 781         else
 782                 ibmveth_set_tx_csum_flags(dev, data);
 783
 784         return rc;
 785 }
 786
 787 static u32 ibmveth_get_rx_csum(struct net_device *dev)
 788 {
 789         struct ibmveth_adapter *adapter = dev->priv;
 790         return adapter->rx_csum;
 791 }
 792
 793 static void ibmveth_get_strings(struct net_device *dev, u32 stringset, u8 *data)
 794 {
 795         int i;
 796
 797         if (stringset != ETH_SS_STATS)
 798                 return;
 799
 800         for (i = 0; i < ARRAY_SIZE(ibmveth_stats); i++, data += ETH_GSTRING_LEN)
 801                 memcpy(data, ibmveth_stats[i].name, ETH_GSTRING_LEN);
 802 }
 803
 804 static int ibmveth_get_sset_count(struct net_device *dev, int sset)
 805 {
 806         switch (sset) {
 807         case ETH_SS_STATS:
 808                 return ARRAY_SIZE(ibmveth_stats);
 809         default:
 810                 return -EOPNOTSUPP;
 811         }
 812 }
 813
 814 static void ibmveth_get_ethtool_stats(struct net_device *dev,
 815                                       struct ethtool_stats *stats, u64 *data)
 816 {
 817         int i;
 818         struct ibmveth_adapter *adapter = dev->priv;
 819
 820         for (i = 0; i < ARRAY_SIZE(ibmveth_stats); i++)
 821                 data[i] = IBMVETH_GET_STAT(adapter, ibmveth_stats[i].offset);
 822 }
 823
 824 static const struct ethtool_ops netdev_ethtool_ops = {
 825         .get_drvinfo            = netdev_get_drvinfo,
 826         .get_settings           = netdev_get_settings,
 827         .get_link               = netdev_get_link,
 828         .set_tx_csum            = ibmveth_set_tx_csum,
 829         .get_rx_csum            = ibmveth_get_rx_csum,
 830         .set_rx_csum            = ibmveth_set_rx_csum,
 831         .get_strings            = ibmveth_get_strings,
 832         .get_sset_count         = ibmveth_get_sset_count,
 833         .get_ethtool_stats      = ibmveth_get_ethtool_stats,
 834 };
 835
 836 static int ibmveth_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
 837 {
 838         return -EOPNOTSUPP;
 839 }
 840
 841 #define page_offset(v) ((unsigned long)(v) & ((1 << 12) - 1))
 842
 843 static int ibmveth_start_xmit(struct sk_buff *skb, struct net_device *netdev)
 844 {
 845         struct ibmveth_adapter *adapter = netdev->priv;
 846         union ibmveth_buf_desc desc;
 847         unsigned long lpar_rc;
 848         unsigned long correlator;
 849         unsigned long flags;
 850         unsigned int retry_count;
 851         unsigned int tx_dropped = 0;
 852         unsigned int tx_bytes = 0;
 853         unsigned int tx_packets = 0;
 854         unsigned int tx_send_failed = 0;
 855         unsigned int tx_map_failed = 0;
 856
 857         desc.fields.flags_len = IBMVETH_BUF_VALID | skb->len;
 858         desc.fields.address = dma_map_single(&adapter->vdev->dev, skb->data,
 859                                              skb->len, DMA_TO_DEVICE);
 860
 861         if (skb->ip_summed == CHECKSUM_PARTIAL &&
 862             ip_hdr(skb)->protocol != IPPROTO_TCP && skb_checksum_help(skb)) {
 863                 ibmveth_error_printk("tx: failed to checksum packet\n");
 864                 tx_dropped++;
 865                 goto out;
 866         }
 867
 868         if (skb->ip_summed == CHECKSUM_PARTIAL) {
 869                 unsigned char *buf = skb_transport_header(skb) + skb->csum_offset;
 870
 871                 desc.fields.flags_len |= (IBMVETH_BUF_NO_CSUM | IBMVETH_BUF_CSUM_GOOD);
 872
 873                 /* Need to zero out the checksum */
 874                 buf[0] = 0;
 875                 buf[1] = 0;
 876         }
 877
 878         if (dma_mapping_error(desc.fields.address)) {
 879                 ibmveth_error_printk("tx: unable to map xmit buffer\n");
 880                 tx_map_failed++;
 881                 tx_dropped++;
 882                 goto out;
 883         }
 884
 885         /* send the frame. Arbitrarily set retrycount to 1024 */
 886         correlator = 0;
 887         retry_count = 1024;
 888         do {
 889                 lpar_rc = h_send_logical_lan(adapter->vdev->unit_address,
 890                                              desc.desc, 0, 0, 0, 0, 0,
 891                                              correlator, &correlator);
 892         } while ((lpar_rc == H_BUSY) && (retry_count--));
 893
 894         if(lpar_rc != H_SUCCESS && lpar_rc != H_DROPPED) {
 895                 ibmveth_error_printk("tx: h_send_logical_lan failed with rc=%ld\n", lpar_rc);
 896                 ibmveth_error_printk("tx: valid=%d, len=%d, address=0x%08x\n",
 897                                      (desc.fields.flags_len & IBMVETH_BUF_VALID) ? 1 : 0,
 898                                      skb->len, desc.fields.address);
 899                 tx_send_failed++;
 900                 tx_dropped++;
 901         } else {
 902                 tx_packets++;
 903                 tx_bytes += skb->len;
 904                 netdev->trans_start = jiffies;
 905         }
 906
 907         dma_unmap_single(&adapter->vdev->dev, desc.fields.address,
 908                          skb->len, DMA_TO_DEVICE);
 909
 910 out:    spin_lock_irqsave(&adapter->stats_lock, flags);
 911         netdev->stats.tx_dropped += tx_dropped;
 912         netdev->stats.tx_bytes += tx_bytes;
 913         netdev->stats.tx_packets += tx_packets;
 914         adapter->tx_send_failed += tx_send_failed;
 915         adapter->tx_map_failed += tx_map_failed;
 916         spin_unlock_irqrestore(&adapter->stats_lock, flags);
 917
 918         dev_kfree_skb(skb);
 919         return 0;
 920 }
 921
 922 static int ibmveth_poll(struct napi_struct *napi, int budget)
 923 {
 924         struct ibmveth_adapter *adapter = container_of(napi, struct ibmveth_adapter, napi);
 925         struct net_device *netdev = adapter->netdev;
 926         int frames_processed = 0;
 927         unsigned long lpar_rc;
 928
 929  restart_poll:
 930         do {
 931                 struct sk_buff *skb;
 932
 933                 if (!ibmveth_rxq_pending_buffer(adapter))
 934                         break;
 935
 936                 rmb();
 937                 if (!ibmveth_rxq_buffer_valid(adapter)) {
 938                         wmb(); /* suggested by larson1 */
 939                         adapter->rx_invalid_buffer++;
 940                         ibmveth_debug_printk("recycling invalid buffer\n");
 941                         ibmveth_rxq_recycle_buffer(adapter);
 942                 } else {
 943                         int length = ibmveth_rxq_frame_length(adapter);
 944                         int offset = ibmveth_rxq_frame_offset(adapter);
 945                         int csum_good = ibmveth_rxq_csum_good(adapter);
 946
 947                         skb = ibmveth_rxq_get_buffer(adapter);
 948                         if (csum_good)
 949                                 skb->ip_summed = CHECKSUM_UNNECESSARY;
 950
 951                         ibmveth_rxq_harvest_buffer(adapter);
 952
 953                         skb_reserve(skb, offset);
 954                         skb_put(skb, length);
 955                         skb->protocol = eth_type_trans(skb, netdev);
 956
 957                         netif_receive_skb(skb); /* send it up */
 958
 959                         netdev->stats.rx_packets++;
 960                         netdev->stats.rx_bytes += length;
 961                         frames_processed++;
 962                         netdev->last_rx = jiffies;
 963                 }
 964         } while (frames_processed < budget);
 965
 966         ibmveth_replenish_task(adapter);
 967
 968         if (frames_processed < budget) {
 969                 /* We think we are done - reenable interrupts,
 970                  * then check once more to make sure we are done.
 971                  */
 972                 lpar_rc = h_vio_signal(adapter->vdev->unit_address,
 973                                        VIO_IRQ_ENABLE);
 974
 975                 ibmveth_assert(lpar_rc == H_SUCCESS);
 976
 977                 netif_rx_complete(netdev, napi);
 978
 979                 if (ibmveth_rxq_pending_buffer(adapter) &&
 980                     netif_rx_reschedule(netdev, napi)) {
 981                         lpar_rc = h_vio_signal(adapter->vdev->unit_address,
 982                                                VIO_IRQ_DISABLE);
 983                         goto restart_poll;
 984                 }
 985         }
 986
 987         return frames_processed;
 988 }
 989
 990 static irqreturn_t ibmveth_interrupt(int irq, void *dev_instance)
 991 {
 992         struct net_device *netdev = dev_instance;
 993         struct ibmveth_adapter *adapter = netdev->priv;
 994         unsigned long lpar_rc;
 995
 996         if (netif_rx_schedule_prep(netdev, &adapter->napi)) {
 997                 lpar_rc = h_vio_signal(adapter->vdev->unit_address,
 998                                        VIO_IRQ_DISABLE);
 999                 ibmveth_assert(lpar_rc == H_SUCCESS);
1000                 __netif_rx_schedule(netdev, &adapter->napi);
1001         }
1002         return IRQ_HANDLED;
1003 }
1004
1005 static void ibmveth_set_multicast_list(struct net_device *netdev)
1006 {
1007         struct ibmveth_adapter *adapter = netdev->priv;
1008         unsigned long lpar_rc;
1009
1010         if((netdev->flags & IFF_PROMISC) || (netdev->mc_count > adapter->mcastFilterSize)) {
1011                 lpar_rc = h_multicast_ctrl(adapter->vdev->unit_address,
1012                                            IbmVethMcastEnableRecv |
1013                                            IbmVethMcastDisableFiltering,
1014                                            0);
1015                 if(lpar_rc != H_SUCCESS) {
1016                         ibmveth_error_printk("h_multicast_ctrl rc=%ld when entering promisc mode\n", lpar_rc);
1017                 }
1018         } else {
1019                 struct dev_mc_list *mclist = netdev->mc_list;
1020                 int i;
1021                 /* clear the filter table & disable filtering */
1022                 lpar_rc = h_multicast_ctrl(adapter->vdev->unit_address,
1023                                            IbmVethMcastEnableRecv |
1024                                            IbmVethMcastDisableFiltering |
1025                                            IbmVethMcastClearFilterTable,
1026                                            0);
1027                 if(lpar_rc != H_SUCCESS) {
1028                         ibmveth_error_printk("h_multicast_ctrl rc=%ld when attempting to clear filter table\n", lpar_rc);
1029                 }
1030                 /* add the addresses to the filter table */
1031                 for(i = 0; i < netdev->mc_count; ++i, mclist = mclist->next) {
1032                         // add the multicast address to the filter table
1033                         unsigned long mcast_addr = 0;
1034                         memcpy(((char *)&mcast_addr)+2, mclist->dmi_addr, 6);
1035                         lpar_rc = h_multicast_ctrl(adapter->vdev->unit_address,
1036                                                    IbmVethMcastAddFilter,
1037                                                    mcast_addr);
1038                         if(lpar_rc != H_SUCCESS) {
1039                                 ibmveth_error_printk("h_multicast_ctrl rc=%ld when adding an entry to the filter table\n", lpar_rc);
1040                         }
1041                 }
1042
1043                 /* re-enable filtering */
1044                 lpar_rc = h_multicast_ctrl(adapter->vdev->unit_address,
1045                                            IbmVethMcastEnableFiltering,
1046                                            0);
1047                 if(lpar_rc != H_SUCCESS) {
1048                         ibmveth_error_printk("h_multicast_ctrl rc=%ld when enabling filtering\n", lpar_rc);
1049                 }
1050         }
1051 }
1052
1053 static int ibmveth_change_mtu(struct net_device *dev, int new_mtu)
1054 {
1055         struct ibmveth_adapter *adapter = dev->priv;
1056         int new_mtu_oh = new_mtu + IBMVETH_BUFF_OH;
1057         int reinit = 0;
1058         int i, rc;
1059
1060         if (new_mtu < IBMVETH_MAX_MTU)
1061                 return -EINVAL;
1062
1063         for (i = 0; i < IbmVethNumBufferPools; i++)
1064                 if (new_mtu_oh < adapter->rx_buff_pool[i].buff_size)
1065                         break;
1066
1067         if (i == IbmVethNumBufferPools)
1068                 return -EINVAL;
1069
1070         /* Look for an active buffer pool that can hold the new MTU */
1071         for(i = 0; i<IbmVethNumBufferPools; i++) {
1072                 if (!adapter->rx_buff_pool[i].active) {
1073                         adapter->rx_buff_pool[i].active = 1;
1074                         reinit = 1;
1075                 }
1076
1077                 if (new_mtu_oh < adapter->rx_buff_pool[i].buff_size) {
1078                         if (reinit && netif_running(adapter->netdev)) {
1079                                 adapter->pool_config = 1;
1080                                 ibmveth_close(adapter->netdev);
1081                                 adapter->pool_config = 0;
1082                                 dev->mtu = new_mtu;
1083                                 if ((rc = ibmveth_open(adapter->netdev)))
1084                                         return rc;
1085                         } else
1086                                 dev->mtu = new_mtu;
1087                         return 0;
1088                 }
1089         }
1090         return -EINVAL;
1091 }
1092
1093 #ifdef CONFIG_NET_POLL_CONTROLLER
1094 static void ibmveth_poll_controller(struct net_device *dev)
1095 {
1096         ibmveth_replenish_task(dev->priv);
1097         ibmveth_interrupt(dev->irq, dev);
1098 }
1099 #endif
1100
1101 static int __devinit ibmveth_probe(struct vio_dev *dev, const struct vio_device_id *id)
1102 {
1103         int rc, i;
1104         long ret;
1105         struct net_device *netdev;
1106         struct ibmveth_adapter *adapter;
1107         u64 set_attr, ret_attr;
1108
1109         unsigned char *mac_addr_p;
1110         unsigned int *mcastFilterSize_p;
1111
1112
1113         ibmveth_debug_printk_no_adapter("entering ibmveth_probe for UA 0x%x\n",
1114                                         dev->unit_address);
1115
1116         mac_addr_p = (unsigned char *) vio_get_attribute(dev,
1117                                                 VETH_MAC_ADDR, NULL);
1118         if(!mac_addr_p) {
1119                 printk(KERN_ERR "(%s:%3.3d) ERROR: Can't find VETH_MAC_ADDR "
1120                                 "attribute\n", __FILE__, __LINE__);
1121                 return 0;
1122         }
1123
1124         mcastFilterSize_p = (unsigned int *) vio_get_attribute(dev,
1125                                                 VETH_MCAST_FILTER_SIZE, NULL);
1126         if(!mcastFilterSize_p) {
1127                 printk(KERN_ERR "(%s:%3.3d) ERROR: Can't find "
1128                                 "VETH_MCAST_FILTER_SIZE attribute\n",
1129                                 __FILE__, __LINE__);
1130                 return 0;
1131         }
1132
1133         netdev = alloc_etherdev(sizeof(struct ibmveth_adapter));
1134
1135         if(!netdev)
1136                 return -ENOMEM;
1137
1138         adapter = netdev->priv;
1139         dev->dev.driver_data = netdev;
1140
1141         adapter->vdev = dev;
1142         adapter->netdev = netdev;
1143         adapter->mcastFilterSize= *mcastFilterSize_p;
1144         adapter->pool_config = 0;
1145
1146         netif_napi_add(netdev, &adapter->napi, ibmveth_poll, 16);
1147
1148         /*      Some older boxes running PHYP non-natively have an OF that
1149                 returns a 8-byte local-mac-address field (and the first
1150                 2 bytes have to be ignored) while newer boxes' OF return
1151                 a 6-byte field. Note that IEEE 1275 specifies that
1152                 local-mac-address must be a 6-byte field.
1153                 The RPA doc specifies that the first byte must be 10b, so
1154                 we'll just look for it to solve this 8 vs. 6 byte field issue */
1155
1156         if ((*mac_addr_p & 0x3) != 0x02)
1157                 mac_addr_p += 2;
1158
1159         adapter->mac_addr = 0;
1160         memcpy(&adapter->mac_addr, mac_addr_p, 6);
1161
1162         netdev->irq = dev->irq;
1163         netdev->open               = ibmveth_open;
1164         netdev->stop               = ibmveth_close;
1165         netdev->hard_start_xmit    = ibmveth_start_xmit;
1166         netdev->set_multicast_list = ibmveth_set_multicast_list;
1167         netdev->do_ioctl           = ibmveth_ioctl;
1168         netdev->ethtool_ops           = &netdev_ethtool_ops;
1169         netdev->change_mtu         = ibmveth_change_mtu;
1170         SET_NETDEV_DEV(netdev, &dev->dev);
1171 #ifdef CONFIG_NET_POLL_CONTROLLER
1172         netdev->poll_controller = ibmveth_poll_controller;
1173 #endif
1174         netdev->features |= NETIF_F_LLTX;
1175         spin_lock_init(&adapter->stats_lock);
1176
1177         memcpy(&netdev->dev_addr, &adapter->mac_addr, netdev->addr_len);
1178
1179         for(i = 0; i<IbmVethNumBufferPools; i++) {
1180                 struct kobject *kobj = &adapter->rx_buff_pool[i].kobj;
1181                 int error;
1182
1183                 ibmveth_init_buffer_pool(&adapter->rx_buff_pool[i], i,
1184                                          pool_count[i], pool_size[i],
1185                                          pool_active[i]);
1186                 error = kobject_init_and_add(kobj, &ktype_veth_pool,
1187                                              &dev->dev.kobj, "pool%d", i);
1188                 if (!error)
1189                         kobject_uevent(kobj, KOBJ_ADD);
1190         }
1191
1192         ibmveth_debug_printk("adapter @ 0x%p\n", adapter);
1193
1194         adapter->buffer_list_dma = DMA_ERROR_CODE;
1195         adapter->filter_list_dma = DMA_ERROR_CODE;
1196         adapter->rx_queue.queue_dma = DMA_ERROR_CODE;
1197
1198         ibmveth_debug_printk("registering netdev...\n");
1199
1200         ret = h_illan_attributes(dev->unit_address, 0, 0, &ret_attr);
1201
1202         if (ret == H_SUCCESS && !(ret_attr & IBMVETH_ILLAN_ACTIVE_TRUNK) &&
1203             !(ret_attr & IBMVETH_ILLAN_TRUNK_PRI_MASK) &&
1204             (ret_attr & IBMVETH_ILLAN_PADDED_PKT_CSUM)) {
1205                 set_attr = IBMVETH_ILLAN_IPV4_TCP_CSUM;
1206
1207                 ret = h_illan_attributes(dev->unit_address, 0, set_attr, &ret_attr);
1208
1209                 if (ret == H_SUCCESS) {
1210                         adapter->rx_csum = 1;
1211                         netdev->features |= NETIF_F_IP_CSUM;
1212                 } else
1213                         ret = h_illan_attributes(dev->unit_address, set_attr, 0, &ret_attr);
1214         }
1215
1216         rc = register_netdev(netdev);
1217
1218         if(rc) {
1219                 ibmveth_debug_printk("failed to register netdev rc=%d\n", rc);
1220                 free_netdev(netdev);
1221                 return rc;
1222         }
1223
1224         ibmveth_debug_printk("registered\n");
1225
1226         ibmveth_proc_register_adapter(adapter);
1227
1228         return 0;
1229 }
1230
1231 static int __devexit ibmveth_remove(struct vio_dev *dev)
1232 {
1233         struct net_device *netdev = dev->dev.driver_data;
1234         struct ibmveth_adapter *adapter = netdev->priv;
1235         int i;
1236
1237         for(i = 0; i<IbmVethNumBufferPools; i++)
1238                 kobject_put(&adapter->rx_buff_pool[i].kobj);
1239
1240         unregister_netdev(netdev);
1241
1242         ibmveth_proc_unregister_adapter(adapter);
1243
1244         free_netdev(netdev);
1245         return 0;
1246 }
1247
1248 #ifdef CONFIG_PROC_FS
1249 static void ibmveth_proc_register_driver(void)
1250 {
1251         ibmveth_proc_dir = proc_mkdir(IBMVETH_PROC_DIR, init_net.proc_net);
1252         if (ibmveth_proc_dir) {
1253         }
1254 }
1255
1256 static void ibmveth_proc_unregister_driver(void)
1257 {
1258         remove_proc_entry(IBMVETH_PROC_DIR, init_net.proc_net);
1259 }
1260
1261 static int ibmveth_show(struct seq_file *seq, void *v)
1262 {
1263         struct ibmveth_adapter *adapter = seq->private;
1264         char *current_mac = ((char*) &adapter->netdev->dev_addr);
1265         char *firmware_mac = ((char*) &adapter->mac_addr) ;
1266         DECLARE_MAC_BUF(mac);
1267
1268         seq_printf(seq, "%s %s\n\n", ibmveth_driver_string, ibmveth_driver_version);
1269
1270         seq_printf(seq, "Unit Address:    0x%x\n", adapter->vdev->unit_address);
1271         seq_printf(seq, "Current MAC:     %s\n", print_mac(mac, current_mac));
1272         seq_printf(seq, "Firmware MAC:    %s\n", print_mac(mac, firmware_mac));
1273
1274         seq_printf(seq, "\nAdapter Statistics:\n");
1275         seq_printf(seq, "  TX:  vio_map_single failres:      %ld\n", adapter->tx_map_failed);
1276         seq_printf(seq, "       send failures:               %ld\n", adapter->tx_send_failed);
1277         seq_printf(seq, "  RX:  replenish task cycles:       %ld\n", adapter->replenish_task_cycles);
1278         seq_printf(seq, "       alloc_skb_failures:          %ld\n", adapter->replenish_no_mem);
1279         seq_printf(seq, "       add buffer failures:         %ld\n", adapter->replenish_add_buff_failure);
1280         seq_printf(seq, "       invalid buffers:             %ld\n", adapter->rx_invalid_buffer);
1281         seq_printf(seq, "       no buffers:                  %ld\n", adapter->rx_no_buffer);
1282
1283         return 0;
1284 }
1285
1286 static int ibmveth_proc_open(struct inode *inode, struct file *file)
1287 {
1288         return single_open(file, ibmveth_show, PDE(inode)->data);
1289 }
1290
1291 static const struct file_operations ibmveth_proc_fops = {
1292         .owner   = THIS_MODULE,
1293         .open    = ibmveth_proc_open,
1294         .read    = seq_read,
1295         .llseek  = seq_lseek,
1296         .release = single_release,
1297 };
1298
1299 static void ibmveth_proc_register_adapter(struct ibmveth_adapter *adapter)
1300 {
1301         struct proc_dir_entry *entry;
1302         if (ibmveth_proc_dir) {
1303                 char u_addr[10];
1304                 sprintf(u_addr, "%x", adapter->vdev->unit_address);
1305                 entry = proc_create_data(u_addr, S_IFREG, ibmveth_proc_dir,
1306                                          &ibmveth_proc_fops, adapter);
1307                 if (!entry)
1308                         ibmveth_error_printk("Cannot create adapter proc entry");
1309         }
1310         return;
1311 }
1312
1313 static void ibmveth_proc_unregister_adapter(struct ibmveth_adapter *adapter)
1314 {
1315         if (ibmveth_proc_dir) {
1316                 char u_addr[10];
1317                 sprintf(u_addr, "%x", adapter->vdev->unit_address);
1318                 remove_proc_entry(u_addr, ibmveth_proc_dir);
1319         }
1320 }
1321
1322 #else /* CONFIG_PROC_FS */
1323 static void ibmveth_proc_register_adapter(struct ibmveth_adapter *adapter)
1324 {
1325 }
1326
1327 static void ibmveth_proc_unregister_adapter(struct ibmveth_adapter *adapter)
1328 {
1329 }
1330 static void ibmveth_proc_register_driver(void)
1331 {
1332 }
1333
1334 static void ibmveth_proc_unregister_driver(void)
1335 {
1336 }
1337 #endif /* CONFIG_PROC_FS */
1338
1339 static struct attribute veth_active_attr;
1340 static struct attribute veth_num_attr;
1341 static struct attribute veth_size_attr;
1342
1343 static ssize_t veth_pool_show(struct kobject * kobj,
1344                               struct attribute * attr, char * buf)
1345 {
1346         struct ibmveth_buff_pool *pool = container_of(kobj,
1347                                                       struct ibmveth_buff_pool,
1348                                                       kobj);
1349
1350         if (attr == &veth_active_attr)
1351                 return sprintf(buf, "%d\n", pool->active);
1352         else if (attr == &veth_num_attr)
1353                 return sprintf(buf, "%d\n", pool->size);
1354         else if (attr == &veth_size_attr)
1355                 return sprintf(buf, "%d\n", pool->buff_size);
1356         return 0;
1357 }
1358
1359 static ssize_t veth_pool_store(struct kobject * kobj, struct attribute * attr,
1360 const char * buf, size_t count)
1361 {
1362         struct ibmveth_buff_pool *pool = container_of(kobj,
1363                                                       struct ibmveth_buff_pool,
1364                                                       kobj);
1365         struct net_device *netdev =
1366             container_of(kobj->parent, struct device, kobj)->driver_data;
1367         struct ibmveth_adapter *adapter = netdev->priv;
1368         long value = simple_strtol(buf, NULL, 10);
1369         long rc;
1370
1371         if (attr == &veth_active_attr) {
1372                 if (value && !pool->active) {
1373                         if (netif_running(netdev)) {
1374                                 if(ibmveth_alloc_buffer_pool(pool)) {
1375                                         ibmveth_error_printk("unable to alloc pool\n");
1376                                         return -ENOMEM;
1377                                 }
1378                                 pool->active = 1;
1379                                 adapter->pool_config = 1;
1380                                 ibmveth_close(netdev);
1381                                 adapter->pool_config = 0;
1382                                 if ((rc = ibmveth_open(netdev)))
1383                                         return rc;
1384                         } else
1385                                 pool->active = 1;
1386                 } else if (!value && pool->active) {
1387                         int mtu = netdev->mtu + IBMVETH_BUFF_OH;
1388                         int i;
1389                         /* Make sure there is a buffer pool with buffers that
1390                            can hold a packet of the size of the MTU */
1391                         for (i = 0; i < IbmVethNumBufferPools; i++) {
1392                                 if (pool == &adapter->rx_buff_pool[i])
1393                                         continue;
1394                                 if (!adapter->rx_buff_pool[i].active)
1395                                         continue;
1396                                 if (mtu <= adapter->rx_buff_pool[i].buff_size)
1397                                         break;
1398                         }
1399
1400                         if (i == IbmVethNumBufferPools) {
1401                                 ibmveth_error_printk("no active pool >= MTU\n");
1402                                 return -EPERM;
1403                         }
1404
1405                         pool->active = 0;
1406                         if (netif_running(netdev)) {
1407                                 adapter->pool_config = 1;
1408                                 ibmveth_close(netdev);
1409                                 adapter->pool_config = 0;
1410                                 if ((rc = ibmveth_open(netdev)))
1411                                         return rc;
1412                         }
1413                 }
1414         } else if (attr == &veth_num_attr) {
1415                 if (value <= 0 || value > IBMVETH_MAX_POOL_COUNT)
1416                         return -EINVAL;
1417                 else {
1418                         if (netif_running(netdev)) {
1419                                 adapter->pool_config = 1;
1420                                 ibmveth_close(netdev);
1421                                 adapter->pool_config = 0;
1422                                 pool->size = value;
1423                                 if ((rc = ibmveth_open(netdev)))
1424                                         return rc;
1425                         } else
1426                                 pool->size = value;
1427                 }
1428         } else if (attr == &veth_size_attr) {
1429                 if (value <= IBMVETH_BUFF_OH || value > IBMVETH_MAX_BUF_SIZE)
1430                         return -EINVAL;
1431                 else {
1432                         if (netif_running(netdev)) {
1433                                 adapter->pool_config = 1;
1434                                 ibmveth_close(netdev);
1435                                 adapter->pool_config = 0;
1436                                 pool->buff_size = value;
1437                                 if ((rc = ibmveth_open(netdev)))
1438                                         return rc;
1439                         } else
1440                                 pool->buff_size = value;
1441                 }
1442         }
1443
1444         /* kick the interrupt handler to allocate/deallocate pools */
1445         ibmveth_interrupt(netdev->irq, netdev);
1446         return count;
1447 }
1448
1449
1450 #define ATTR(_name, _mode)      \
1451         struct attribute veth_##_name##_attr = {               \
1452         .name = __stringify(_name), .mode = _mode, \
1453         };
1454
1455 static ATTR(active, 0644);
1456 static ATTR(num, 0644);
1457 static ATTR(size, 0644);
1458
1459 static struct attribute * veth_pool_attrs[] = {
1460         &veth_active_attr,
1461         &veth_num_attr,
1462         &veth_size_attr,
1463         NULL,
1464 };
1465
1466 static struct sysfs_ops veth_pool_ops = {
1467         .show   = veth_pool_show,
1468         .store  = veth_pool_store,
1469 };
1470
1471 static struct kobj_type ktype_veth_pool = {
1472         .release        = NULL,
1473         .sysfs_ops      = &veth_pool_ops,
1474         .default_attrs  = veth_pool_attrs,
1475 };
1476
1477
1478 static struct vio_device_id ibmveth_device_table[] __devinitdata= {
1479         { "network", "IBM,l-lan"},
1480         { "", "" }
1481 };
1482 MODULE_DEVICE_TABLE(vio, ibmveth_device_table);
1483
1484 static struct vio_driver ibmveth_driver = {
1485         .id_table       = ibmveth_device_table,
1486         .probe          = ibmveth_probe,
1487         .remove         = ibmveth_remove,
1488         .driver         = {
1489                 .name   = ibmveth_driver_name,
1490                 .owner  = THIS_MODULE,
1491         }
1492 };
1493
1494 static int __init ibmveth_module_init(void)
1495 {
1496         ibmveth_printk("%s: %s %s\n", ibmveth_driver_name, ibmveth_driver_string, ibmveth_driver_version);
1497
1498         ibmveth_proc_register_driver();
1499
1500         return vio_register_driver(&ibmveth_driver);
1501 }
1502
1503 static void __exit ibmveth_module_exit(void)
1504 {
1505         vio_unregister_driver(&ibmveth_driver);
1506         ibmveth_proc_unregister_driver();
1507 }
1508
1509 module_init(ibmveth_module_init);
1510 module_exit(ibmveth_module_exit);